# LMIs in Control/pages/Delay Dependent Time-Delay Stabilization

Stabilization of Time-Delay Systems - Delay Dependent Case

Suppose, for instance, there was a system where a time-delay was introduced. In that instance, stabilization would have to be done in a different manner. The following example demonstrates how one can stabilize such a system while being dependent on the delay.

## The System

For this particular problem, suppose that we were given the time-delayed system in the form of:

{\displaystyle {\begin{aligned}{\begin{cases}{\dot {x}}(t)&=Ax(t)+A_{d}(t-d)+Bu(t),\\x(t)&=\phi (t),t\in [0,d],0

where

{\displaystyle {\begin{aligned}&A{\text{, }}{A_{d}}\in \mathbb {R} ^{n\times n}{\text{, }}{B}\in \mathbb {R} ^{n\times r}{\text{ are the system coefficient matrices,}}\\&\phi (t){\text{ is the initial condition,}}\\&d{\text{ represents the time-delay, and}}\\&{\bar {d}}{\text{ is a known upper-bound of }}d\\\end{aligned}}}

Then the LMI for determining the Time-Delay System for the Delay-Dependent case would be obtained as described below.

## The Data

In order to obtain the LMI, we need the following 3 matrices: ${\displaystyle A{\text{, }}{A_{d}}{\text{, and }}B}$ .

## The Optimization Problem

Suppose - for the time-delayed system given above - we were asked to design a memoryless feedback control law where ${\displaystyle u(t)=Kx(t)}$  such that the closed-loop system:

{\displaystyle {\begin{aligned}{\begin{cases}{\dot {x}}(t)&=(A+BK)x(t)+A_{d}(t-d),\\x(t)&=\phi (t),t\in [0,d],0

is simultaneously both uniform and asymptotically stable, then the system would be stabilized as follows.

## The LMI: The Delay-Dependent Stabilization of Time-Delay Systems

From the given pieces of information, it is clear that the optimization problem only has a solution if there exists a scalar ${\displaystyle 0<\beta <1}$ , a symmetric matrix ${\displaystyle X>0}$  and a matrix ${\displaystyle W}$  that satisfy the following:

{\displaystyle {\begin{aligned}{\begin{bmatrix}\Phi (X,W)&&{\bar {d}}(X{A^{T}}+{W^{T}}{B^{T}})&&{\bar {d}}X{A_{d}^{T}}\\{\bar {d}}(AX+BW)&&-{\bar {d}}{\beta }I&&0\\{\bar {d}}{A_{d}}X&&0&&-{\bar {d}}(1-\beta )I\end{bmatrix}}&<0\\\end{aligned}}}

where${\displaystyle \Phi (X,W)=X{(A+{A_{d}})^{T}}+(A+{A_{d}})X+BW+{W^{T}}{B^{T}}+{\bar {d}}{A_{d}}{A_{d}^{T}}}$

## Conclusion:

Given the resulting stabilizing control gain matrix ${\displaystyle K=WX^{-1}}$ , it can be observed that the matrix is asymptotically stable from the time-delay system where this gain matrix was derived.

## Implementation

• Example Code - A GitHub link that contains code (titled "DelayDependentTimeDelay.m") that demonstrates how this LMI can be implemented using MATLAB-YALMIP.